Skip to content

GitLab

Switch branch/tag
  1. 03 Nov, 2023 3 commits
  3. 02 Nov, 2023 6 commits
  5. 01 Nov, 2023 7 commits
  7. 31 Oct, 2023 7 commits
  9. 30 Oct, 2023 6 commits
  11. 27 Oct, 2023 3 commits
    • Patrick von Platen's avatar
      [Attention Mask] Refactor all encoder-decoder attention mask (#27086) · ac589375
      Patrick von Platen authored 
      

* [FA2 Bart] Add FA2 to all Bart-like

* better

* Refactor attention mask

* remove all customized atteniton logic

* format

* mass rename

* replace _expand_mask

* replace _expand_mask

* mass rename

* add pt files

* mass replace & rename

* mass replace & rename

* mass replace & rename

* mass replace & rename

* Update src/transformers/models/idefics/modeling_idefics.py

* fix more

* clean more

* fix more

* make style

* fix again

* finish

* finish

* finish

* finish

* finish

* finish

* finish

* finish

* finish

* finish

* Apply suggestions from code review

* Apply suggestions from code review
Co-authored-by: default avataramyeroberts <22614925+amyeroberts@users.noreply.github.com>

* small fix mistral

* finish

* finish

* finish

* finish

---------
Co-authored-by: default avataramyeroberts <22614925+amyeroberts@users.noreply.github.com>
      ac589375
    • Isaac Chung's avatar
      Add early stopping for Bark generation via logits processor (#26675) · e2bffcfa
      Isaac Chung authored 
      * add early stopping logits processor

* black formmated

* indent

* follow method signature

* actual logic

* check for None

* address comments on docstrings and method signature

* add unit test under `LogitsProcessorTest` wip

* unit test passing

* black formatted

* condition per sample

* add to BarkModelIntegrationTests

* wip BarkSemanticModelTest

* rename and add to kwargs handling

* not add to BarkSemanticModelTest

* correct logic and assert last outputs tokens different in test

* doc-builder style

* read from kwargs as well

* assert len of with less than that of without

* ruff

* add back seed and test case

* add original impl default suggestion

* doc-builder

* rename and use softmax

* switch back to LogitsProcessor and update docs wording

* camelCase and spelling and saving compute

* assert strictly less than

* assert less than

* expand test_generate_semantic_early_stop instead
      e2bffcfa
    • Arthur's avatar
      Revert "add exllamav2 arg" (#27102) · 90ee9cea
      Arthur authored 
      Revert "add exllamav2 arg (#26437)"

This reverts commit 8214d6e7.
      90ee9cea
  13. 26 Oct, 2023 3 commits
    • Zach Mueller's avatar
      Save TB logs as part of push_to_hub (#27022) · 34a64064
      Zach Mueller authored 
      * Support runs/

* Upload runs folder as part of push to hub

* Add a test

* Add to test deps

* Update with proposed solution from Slack

* Ensure that repo gets deleted in tests
      34a64064
    • Marc Sun's avatar
      add exllamav2 arg (#26437) · 8214d6e7
      Marc Sun authored 
      * add_ xllamav2 arg

* add test

* style

* add check

* add doc

* replace by use_exllama_v2

* fix tests

* fix doc

* style

* better condition

* fix logic

* add deprecate msg
      8214d6e7
    • Arthur's avatar
      Add-support for commit description (#26704) · 4864d08d
      Arthur authored 
      * fix

* update

* revert

* add dosctring

* good to go

* update

* add a test
      4864d08d
  15. 25 Oct, 2023 2 commits
  17. 24 Oct, 2023 3 commits
    • JB (Don)'s avatar
      Add a default decoder_attention_mask for EncoderDecoderModel during training (#26752) · a0fd3448
      JB (Don) authored 
      * Add a default decoder_attention_mask for EncoderDecoderModel during training

Since we are already creating the default decoder_input_ids from the labels, we should also
create a default decoder_attention_mask to go with it.

* Fix test constant that relied on manual_seed()

The test was changed to use a decoder_attention_mask that ignores padding instead (which is
the default one created by BERT when attention_mask is None).

* Create the decoder_attention_mask using decoder_input_ids instead of labels

* Fix formatting in test
      a0fd3448
    • Alex McKinney's avatar
      Device agnostic testing (#25870) · 9da45171
      Alex McKinney authored 
      

* adds agnostic decorators and availability fns

* renaming decorators and fixing imports

* updating some representative example tests
bloom, opt, and reformer for now

* wip device agnostic functions

* lru cache to device checking functions

* adds `TRANSFORMERS_TEST_DEVICE_SPEC`
if present, imports the target file and updates device to function
mappings

* comments `TRANSFORMERS_TEST_DEVICE_SPEC` code

* extra checks on device name

* `make style; make quality`

* updates default functions for agnostic calls

* applies suggestions from review

* adds `is_torch_available` guard

* Add spec file to docs, rename function dispatch names to backend_*

* add backend import to docs example for spec file

* change instances of  to

* Move register backend to before device check as per @statelesshz changes

* make style

* make opt test require fp16 to run

---------
Co-authored-by: default avatararsalanu <arsalanu@graphcore.ai>
Co-authored-by: default avatararsalanu <hzji210@gmail.com>
      9da45171
    • Xuehai Pan's avatar
      Register ModelOutput as supported torch pytree nodes (#26618) · cc7803c0
      Xuehai Pan authored 
      * Register ModelOutput as supported torch pytree nodes

* Test ModelOutput as supported torch pytree nodes

* Update type hints for pytree unflatten functions
      cc7803c0